OpenViking is an agent-native context database that unifies memories, resources, and skills under a virtual filesystem paradigm (viking:// URIs). It provides a three-tier context model (L0 abstract, L1 overview, L2 full content), recursive directory-based retrieval with visualized trajectories, and automatic session memory extraction. The system consists of a FastAPI HTTP server with embedded Rust components (RAGFS filesystem, vector index engine), a Rust CLI (ov), Python SDKs, a React/Vite Web Studio dashboard, and VikingBot - an agent framework with multi-channel support (Telegram, Slack, Feishu, etc.) and MCP integration. It integrates with OpenClaw, Claude Code, Codex, OpenCode, Hermes, and LangChain/LangGraph. The codebase demonstrates strong engineering: clean separation of concerns, async-first design, comprehensive test coverage (unit, integration, e2e), and production-ready deployment artifacts (Docker, Helm, multi-arch builds). Benchmarks show significant improvements over baseline agents on LoCoMo (82% vs 24-57% accuracy), tau2-bench (+6-12pp), and HotpotQA (91% at top-20).
AutoCoder Nano is an AI-powered coding assistant that leverages large language models (LLMs) for code generation, understanding, and modification. It provides both a command-line interface (CLI) and a web-based gateway server for interactive coding support. The system enables users to generate code, edit files, and answer questions about their codebase through natural language interactions with LLMs. The architecture is modular, consisting of a core LLM client for provider interactions, an agent-based system for task execution, context management for conversation history and file indexing, a retrieval-augmented generation (RAG) mechanism for relevant code retrieval, and a code editing module using SEARCH/REPLACE blocks. Additional components include a skill-based extension system, file locking utilities, and configuration management. The system supports multiple LLM providers (OpenAI, Volcengine, etc.) and integrates with Git for version control operations. Key entry points are the CLI (`auto_coder_nano.py`) and the gateway server (`auto_coder_gateway.py`). Data flows involve user input processed by the agent system, which coordinates LLM interactions, tool usage (file operations, web search, RAG), and context updates. External dependencies include FastAPI, DuckDB with VSS extension, HuggingFace tokenizers, and various utility libraries. The project demonstrates active maintenance with recent commits and exhibits strong engineering practices in its modular design and reusable patterns.
DeepV-Ki is an AI-powered wiki generator for code repositories that combines a Next.js frontend with a FastAPI backend. The system allows users to input a repository URL and automatically generates documentation, architecture diagrams (Mermaid), and provides a RAG-based Q&A interface. It features multi-LLM provider support (OpenAI, Google Gemini, Azure, AWS Bedrock, Ollama), GitLab/GitHub/Bitbucket integration, asynchronous task processing with Celery, and multi-layer caching. The architecture follows a monorepo structure with clear separation between frontend (React/Next.js) and backend (Python/FastAPI) components, communicating via REST APIs. The system includes sophisticated authentication (SSO, session management), progress tracking, and real-time WebSocket communication for streaming AI responses.